Anthropic CEO says open-weight models are not the threat; warns of Chinese AI risks

Anthropic founder and CEO Dario Amodei said the company has never advocated for a ban on open-weight models, directly addressing industry speculation that Anthropic supports broad restrictions on open-weight AI systems.

Amodei argued that open-weight models without dangerous capabilities function as a public good, providing value to businesses, developers, and researchers at minimal cost beyond compute.

He distinguished this view from his concerns about authoritarian governments—particularly the Chinese Communist Party—developing more powerful AI for purposes such as permanent military superiority or domestic repression.

Amodei also warned of AI-enabled biological attacks, asserting that open-weight models are harder to guardrail or withdraw once released, citing a UK AI Security Institute report that once open-weights are released they cannot be withdrawn.

He reiterated support for restricting China’s access to powerful chips, a longstanding U.S. policy, and called for a formal crackdown on model distillation practices.

Amodei expressed support for global model safety testing initiatives, including those led by the U.S., provided such testing is global in scope and includes China, and would apply to the most capable models regardless of origin or openness.

He noted recent U.S. government movement toward such testing and industry proposals that exempt less capable models, while emphasizing that effectiveness would require participation from China.